System Architecture Simulation

Low-Bandwidth Audio Extraction & AI Translation Pipeline (School Project Model)

1
User Device (Video Buffering)
Video is stuck buffering due to slow internet, but audio is extracted immediately.
Video Frame Load: 0%
Audio Extractor:
Extracting Track...
2
Low-Bandwidth Transmission
Audio is sent in tiny chunks (kilobytes) over the network before the video loads.
3
AI NLP Server (Anti-Hallucination)
AI listens to audio chunks, filters errors, and outputs accurate Sinhala text.
[System] Listening to stream...
4
Vector Normalization Engine (Anti-Stretching)
Converts signs into 3D Vectors (X,Y,Z) & Angles for Hands, Fingers, and Face. This stops the avatar from stretching when trained on different people's videos.
WordPose (X,Y,Z)Fingers/FaceStatus
---Waiting
5
VPS Server - TOON Protocol Stream
Instead of heavy JSON, we send TOON (Token Oriented Object Notation) strings back to the phone.
6
EduSign Engine (Anti-Glitch Spline Interpolation)
Applies Kalman Filters & Spline Interpolation to the incoming TOON vectors. This fills in the missing data between frames, ensuring the avatar moves 100% smoothly without any jittering or glitches!
Raw Frames: Jitter Detected
Kalman Filter: Noise Removed
Spline Engine: Smooth Motion Active